The Week Forward - 5 Jul 26

To get prepared for the trading week ahead, i always found it particularly useful to organize my thoughts and to prepare a battle plan prior.
*Recap of Last Week
A quick recap on last week’s price action. We had a huge momentum sell off last week, reaching the most brutal 2day move since Covid. It was not a semis sell off, rather it was a factor sell off ($NVDA held firm while memory puked nearly 20%). In terms of positioning, i believe that most of the crowded longs and “weak hands” have been washed out. For context, pod funds have been implicitly long momentum in this whole wave. These funds also operate on extremely tight drawdown limits. Hence compounding moves especially on the downside. Thursday’s close was extremely strong (for the prior price action), which leads me to believe that the deleveraging is done. We saw a repeat of this into Korea and JP market trading on Friday with $SKYHYNIX and $285A both being sold into the open before staging a massive rally.
As a result, I am calling for this momentum sell off to be over. I see that positioning overhang as being cleared and the market ready to resume its course.
Additionally, on thursday we also saw a broadening out of the rally in other sectors. Should semis and momentum names stage a mean reversal bounce, we will see ATH on the indicies.
This will quickly prompt a rerisking wave and bids to mechanically push up semis further.
*META
On fundamentals, we saw $META’s AI chief Alexandr Wang refuting that META is giving up on their own models. Instead he says “Our next Muse Spark update is coming soon. Big improvements in coding and agentic capabilities to be more competitive with other leading models”
Also noting that the upcoming AI model codenamed “Watermelon, has caught up with OpenAI’s flagship GPT 5.5 model” and “Watermelon uses an order of magnitude more compute than Avocado”
I suspect that with the 3 day weekend to properly digest news, funds who sold off on the Meta news and prior price action will start to bid back exposure. Especially if price starts to move back up again.
My first initial read on the Meta news was that Zuck was in fact going to increase Capex. Note that the hyperscalers treat AI as an existential race. Losing the race means going out of business in the future. Those comments referenced on the Bloomberg article were probably a way for Zuck to reassure investors that they will always be able to find a way to monetize the compute and that overbuilding was not an issue.
*Calender
There is no notable macro news release slated for this week.
On the 10th of July, SK Hynix will debut on the NDX under the symbol $SKHY. I believe that this will be a catalyst for memory and especially $SKHYNIX to rally into.
*On Neoclouds
Last week’s price action may turn into a boon for Neoclouds, by bringing attention into this widely misunderstood and mispriced sector. First, let me lay out the undeniable fact - compute is in a shortage. Dont believe me? Try renting out some GPUs.
The payback period for GPU rental is about 2 -4 years depending on what price you use on your revenue.
For neoclouds, the buildout is derisked by the fact that they get the customers first. The customer prepays even.
Why does this fantastic business exist? Why do the hyperscalers rent from the neoclouds? Why dont the hyperscalers just buy all the GPUs from NVDA?
Because NVDA is playing the long game here by encouraging customer diversity. If the hyperscalers become the only clients, NVDA becomes locked in. By encouraging the growth of neoclouds, NVDA is locking in the hyperscalers instead.
This is why NVDA supports the growth of neoclouds.
The 2 biggest bear cases for Neoclouds is depreciation and the financing.
In a buildout, GPU accounts for ~40% of the cost. This week, NVDA has just announced its strategic partnership program where it assists in the financing for the GPU purchase. This is extremely positive for neoclouds because it lowers the cost of financing (NVDA credit rating vs their own) and also allows for neoclouds to source for a wider range of clients. Previously, neos would rely more heavily on the hyperscalers who had better credit than smaller firms (who might pay more).
I am expecting the client base for the neoclouds to widen. Hyperscalers who could rely on their credit rating to lower pricing will find that edge being challenged away. For example, new neoclouds like $SHAZ, no longer need to find and sign a hyperscaler as an anchor client to secure financing.
On pricing, remember that SPCX is getting 920M a month from GOOG for 110k of NVDA GPUs (GB200) - this works out to be about 11.6/hour
*DEEPSEEK
Deepseek is introducing surge pricing - during peek hours, API prices for V4 Pro and V4 flash will double.
Even the cheap chinese AI is hiking prices now…
*THOUGHTS ON THE WEEK
From this point, i am expecting
1. Momentum sell off to be over
2. For $SKHYNIX to outperform into the ADR listing on 10th Jul
3. For neoclouds to stage a massive bounce. NBIS to break 300 during this month
4. New ATHs on SPX and NDX